[PATCH v9 25/25] dept: Track the potential waits of PG_{locked,writeback}

From: Byungchul Park
Date: Tue Jan 31 2023 - 03:44:35 EST

Next message: Muhammad Usama Anjum: "Re: [PATCH v8 1/4] userfaultfd: Add UFFD WP Async support"
Previous message: Byungchul Park: "[PATCH v9 23/25] dept: Record the latest one out of consecutive waits of the same class"
In reply to: Byungchul Park: "[PATCH v9 23/25] dept: Record the latest one out of consecutive waits of the same class"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Currently, Dept only tracks the real waits of PG_{locked,writeback} that
actually happened having gone through __schedule() to avoid false
positives. However, it ends in limited capacity for deadlock detection,
because anyway there might be still way more potential dependencies by
the waits that have yet to happen but may happen in the future so as to
cause a deadlock.

So let Dept assume that when PG_{locked,writeback} bit gets cleared,
there might be waits on the bit to be woken up.

Even though false positives may increase with the aggressive tracking,
it's worth doing it because it's going to be useful in practice. See the
following link for instance:

https://lore.kernel.org/lkml/1674268856-31807-1-git-send-email-byungchul.park@xxxxxxx/

Signed-off-by: Byungchul Park <max.byungchul.park@xxxxxxxxx>
---
include/linux/mm_types.h | 3 ++
include/linux/page-flags.h | 112 ++++++++++++++++++++++++++++++++++++++++-----
include/linux/pagemap.h | 7 ++-
mm/filemap.c | 11 ++++-
mm/page_alloc.c | 3 ++
5 files changed, 121 insertions(+), 15 deletions(-)

diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 3b84750..61d982e 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -19,6 +19,7 @@
#include <linux/workqueue.h>
#include <linux/seqlock.h>
#include <linux/percpu_counter.h>
+#include <linux/dept.h>

#include <asm/mmu.h>

@@ -252,6 +253,8 @@ struct page {
#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
int _last_cpupid;
#endif
+ struct dept_ext_wgen PG_locked_wgen;
+ struct dept_ext_wgen PG_writeback_wgen;
} _struct_page_alignment;

/*
diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
index 69e93a0..d6ca114 100644
--- a/include/linux/page-flags.h
+++ b/include/linux/page-flags.h
@@ -202,6 +202,50 @@ enum pageflags {

#ifndef __GENERATING_BOUNDS_H

+#ifdef CONFIG_DEPT
+#include <linux/kernel.h>
+#include <linux/dept.h>
+
+extern struct dept_map PG_locked_map;
+extern struct dept_map PG_writeback_map;
+
+/*
+ * Place the following annotations in its suitable point in code:
+ *
+ * Annotate dept_page_set_bit() around firstly set_bit*()
+ * Annotate dept_page_clear_bit() around clear_bit*()
+ * Annotate dept_page_wait_on_bit() around wait_on_bit*()
+ */
+
+static inline void dept_page_set_bit(struct page *p, int bit_nr)
+{
+ if (bit_nr == PG_locked)
+ dept_request_event(&PG_locked_map, &p->PG_locked_wgen);
+ else if (bit_nr == PG_writeback)
+ dept_request_event(&PG_writeback_map, &p->PG_writeback_wgen);
+}
+
+static inline void dept_page_clear_bit(struct page *p, int bit_nr)
+{
+ if (bit_nr == PG_locked)
+ dept_event(&PG_locked_map, 1UL, _RET_IP_, __func__, &p->PG_locked_wgen);
+ else if (bit_nr == PG_writeback)
+ dept_event(&PG_writeback_map, 1UL, _RET_IP_, __func__, &p->PG_writeback_wgen);
+}
+
+static inline void dept_page_wait_on_bit(struct page *p, int bit_nr)
+{
+ if (bit_nr == PG_locked)
+ dept_wait(&PG_locked_map, 1UL, _RET_IP_, __func__, 0, -1L);
+ else if (bit_nr == PG_writeback)
+ dept_wait(&PG_writeback_map, 1UL, _RET_IP_, __func__, 0, -1L);
+}
+#else
+#define dept_page_set_bit(p, bit_nr) do { } while (0)
+#define dept_page_clear_bit(p, bit_nr) do { } while (0)
+#define dept_page_wait_on_bit(p, bit_nr) do { } while (0)
+#endif
+
#ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
DECLARE_STATIC_KEY_FALSE(hugetlb_optimize_vmemmap_key);

@@ -383,44 +427,88 @@ static unsigned long *folio_flags(struct folio *folio, unsigned n)
#define SETPAGEFLAG(uname, lname, policy) \
static __always_inline \
void folio_set_##lname(struct folio *folio) \
-{ set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
+{ \
+ set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); \
+ dept_page_set_bit(&folio->page, PG_##lname); \
+} \
static __always_inline void SetPage##uname(struct page *page) \
-{ set_bit(PG_##lname, &policy(page, 1)->flags); }
+{ \
+ set_bit(PG_##lname, &policy(page, 1)->flags); \
+ dept_page_set_bit(page, PG_##lname); \
+}

#define CLEARPAGEFLAG(uname, lname, policy) \
static __always_inline \
void folio_clear_##lname(struct folio *folio) \
-{ clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
+{ \
+ clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); \
+ dept_page_clear_bit(&folio->page, PG_##lname); \
+} \
static __always_inline void ClearPage##uname(struct page *page) \
-{ clear_bit(PG_##lname, &policy(page, 1)->flags); }
+{ \
+ clear_bit(PG_##lname, &policy(page, 1)->flags); \
+ dept_page_clear_bit(page, PG_##lname); \
+}

#define __SETPAGEFLAG(uname, lname, policy) \
static __always_inline \
void __folio_set_##lname(struct folio *folio) \
-{ __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
+{ \
+ __set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); \
+ dept_page_set_bit(&folio->page, PG_##lname); \
+} \
static __always_inline void __SetPage##uname(struct page *page) \
-{ __set_bit(PG_##lname, &policy(page, 1)->flags); }
+{ \
+ __set_bit(PG_##lname, &policy(page, 1)->flags); \
+ dept_page_set_bit(page, PG_##lname); \
+}

#define __CLEARPAGEFLAG(uname, lname, policy) \
static __always_inline \
void __folio_clear_##lname(struct folio *folio) \
-{ __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
+{ \
+ __clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); \
+ dept_page_clear_bit(&folio->page, PG_##lname); \
+} \
static __always_inline void __ClearPage##uname(struct page *page) \
-{ __clear_bit(PG_##lname, &policy(page, 1)->flags); }
+{ \
+ __clear_bit(PG_##lname, &policy(page, 1)->flags); \
+ dept_page_clear_bit(page, PG_##lname); \
+}

#define TESTSETFLAG(uname, lname, policy) \
static __always_inline \
bool folio_test_set_##lname(struct folio *folio) \
-{ return test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
+{ \
+ bool ret = test_and_set_bit(PG_##lname, folio_flags(folio, FOLIO_##policy));\
+ if (!ret) \
+ dept_page_set_bit(&folio->page, PG_##lname); \
+ return ret; \
+} \
static __always_inline int TestSetPage##uname(struct page *page) \
-{ return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
+{ \
+ bool ret = test_and_set_bit(PG_##lname, &policy(page, 1)->flags);\
+ if (!ret) \
+ dept_page_set_bit(page, PG_##lname); \
+ return ret; \
+}

#define TESTCLEARFLAG(uname, lname, policy) \
static __always_inline \
bool folio_test_clear_##lname(struct folio *folio) \
-{ return test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy)); } \
+{ \
+ bool ret = test_and_clear_bit(PG_##lname, folio_flags(folio, FOLIO_##policy));\
+ if (ret) \
+ dept_page_clear_bit(&folio->page, PG_##lname); \
+ return ret; \
+} \
static __always_inline int TestClearPage##uname(struct page *page) \
-{ return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
+{ \
+ bool ret = test_and_clear_bit(PG_##lname, &policy(page, 1)->flags);\
+ if (ret) \
+ dept_page_clear_bit(page, PG_##lname); \
+ return ret; \
+}

#define PAGEFLAG(uname, lname, policy) \
TESTPAGEFLAG(uname, lname, policy) \
diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 29e1f9e..2843619 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -898,7 +898,12 @@ bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
*/
static inline bool folio_trylock(struct folio *folio)
{
- return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
+ bool ret = !test_and_set_bit_lock(PG_locked, folio_flags(folio, 0));
+
+ if (ret)
+ dept_page_set_bit(&folio->page, PG_locked);
+
+ return likely(ret);
}

/*
diff --git a/mm/filemap.c b/mm/filemap.c
index adc49cb..b80c8e2 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1097,6 +1097,7 @@ static int wake_page_function(wait_queue_entry_t *wait, unsigned mode, int sync,
if (flags & WQ_FLAG_CUSTOM) {
if (test_and_set_bit(key->bit_nr, &key->folio->flags))
return -1;
+ dept_page_set_bit(&key->folio->page, key->bit_nr);
flags |= WQ_FLAG_DONE;
}
}
@@ -1206,6 +1207,7 @@ static inline bool folio_trylock_flag(struct folio *folio, int bit_nr,
if (wait->flags & WQ_FLAG_EXCLUSIVE) {
if (test_and_set_bit(bit_nr, &folio->flags))
return false;
+ dept_page_set_bit(&folio->page, bit_nr);
} else if (test_bit(bit_nr, &folio->flags))
return false;

@@ -1216,8 +1218,10 @@ static inline bool folio_trylock_flag(struct folio *folio, int bit_nr,
/* How many times do we accept lock stealing from under a waiter? */
int sysctl_page_lock_unfairness = 5;

-static struct dept_map __maybe_unused PG_locked_map = DEPT_MAP_INITIALIZER(PG_locked_map, NULL);
-static struct dept_map __maybe_unused PG_writeback_map = DEPT_MAP_INITIALIZER(PG_writeback_map, NULL);
+struct dept_map __maybe_unused PG_locked_map = DEPT_MAP_INITIALIZER(PG_locked_map, NULL);
+struct dept_map __maybe_unused PG_writeback_map = DEPT_MAP_INITIALIZER(PG_writeback_map, NULL);
+EXPORT_SYMBOL(PG_locked_map);
+EXPORT_SYMBOL(PG_writeback_map);

static inline int folio_wait_bit_common(struct folio *folio, int bit_nr,
int state, enum behavior behavior)
@@ -1230,6 +1234,7 @@ static inline int folio_wait_bit_common(struct folio *folio, int bit_nr,
unsigned long pflags;
bool in_thrashing;

+ dept_page_wait_on_bit(&folio->page, bit_nr);
if (bit_nr == PG_locked)
sdt_might_sleep_start(&PG_locked_map);
else if (bit_nr == PG_writeback)
@@ -1327,6 +1332,7 @@ static inline int folio_wait_bit_common(struct folio *folio, int bit_nr,
wait->flags |= WQ_FLAG_DONE;
break;
}
+ dept_page_set_bit(&folio->page, bit_nr);

/*
* If a signal happened, this 'finish_wait()' may remove the last
@@ -1534,6 +1540,7 @@ void folio_unlock(struct folio *folio)
BUILD_BUG_ON(PG_waiters != 7);
BUILD_BUG_ON(PG_locked > 7);
VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
+ dept_page_clear_bit(&folio->page, PG_locked);
if (clear_bit_unlock_is_negative_byte(PG_locked, folio_flags(folio, 0)))
folio_wake_bit(folio, PG_locked);
}
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 0745aed..57d6c82 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -76,6 +76,7 @@
#include <linux/khugepaged.h>
#include <linux/buffer_head.h>
#include <linux/delayacct.h>
+#include <linux/dept.h>
#include <asm/sections.h>
#include <asm/tlbflush.h>
#include <asm/div64.h>
@@ -1626,6 +1627,8 @@ static void __meminit __init_single_page(struct page *page, unsigned long pfn,
page_mapcount_reset(page);
page_cpupid_reset_last(page);
page_kasan_tag_reset(page);
+ dept_ext_wgen_init(&page->PG_locked_wgen);
+ dept_ext_wgen_init(&page->PG_writeback_wgen);

INIT_LIST_HEAD(&page->lru);
#ifdef WANT_PAGE_VIRTUAL
--
1.9.1

Next message: Muhammad Usama Anjum: "Re: [PATCH v8 1/4] userfaultfd: Add UFFD WP Async support"
Previous message: Byungchul Park: "[PATCH v9 23/25] dept: Record the latest one out of consecutive waits of the same class"
In reply to: Byungchul Park: "[PATCH v9 23/25] dept: Record the latest one out of consecutive waits of the same class"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]